ChatOllama won't use with_fallbacks when I use astream_events.

[gbaian10]

Checked other resources

• I added a very descriptive title to this issue.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.
• I am sure that this is a bug in LangChain rather than my code.
• The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).

Example Code

import asyncio
from enum import Enum

from dotenv import load_dotenv
from langchain_core.output_parsers import PydanticToolsParser
from langchain_ollama import ChatOllama
from langchain_openai import ChatOpenAI
from pydantic.v1 import BaseModel, Field

load_dotenv()


class DateEnum(str, Enum):
    first_day = "2024-10-10 10:00:00"
    second_day = "2024-10-11 14:00:00"
    third_day = "2024-10-12 14:00:00"


class SelectItem(BaseModel):
    """Confirm the user's choice based on the user's answer."""

    item: DateEnum = Field(..., description="Select a date based on user responses")


tools = [SelectItem]
ollama_llm = ChatOllama(model="llama3.1:8b").bind_tools(tools)
openai_llm = ChatOpenAI(model="gpt-4o-mini").bind_tools(tools)
parser = PydanticToolsParser(tools=tools)

chain = ollama_llm | parser
fall_back_chain = openai_llm | parser
with_fallback_chain = chain.with_fallbacks([fall_back_chain])

messages = [
    ("ai", f"Which day is most convenient for you in {list(DateEnum)}?"),
    ("human", "30"),
]


async def main():
    async for event in with_fallback_chain.astream_events(messages, version="v2"):
        print(event)  # It will not call fall_back
    print("-" * 20)
    print(await with_fallback_chain.ainvoke(messages))  # It will call fall_back


asyncio.run(main())

Error Message and Stack Trace (if applicable)

Description

ChatOllama won't use with_fallbacks when I use astream_events.
But it will use with_fallbacks when I use ainvoke.

My goal is to know which model produced this output.
When I connect PydanticToolsParser behind the model output, I can't seem to know who generated it. (it is hidden in the AIMessage of the intermediate model output).

So I wanted to take out the intermediate result from astream_events to determine who generated it.
Later I found that ChatOllama seems to be unable to call fall_back under astream_events? Is there a better solution?

System Info

langchain==0.2.11
langchain-core==0.2.24
langchain-ollama==0.1.0
langchain-openai==0.1.19

platform linux
python version = 3.10.12

[ji24077]

Hi, I am ji, 4th year student at UofT CS. I am working with @Ismail-Bashir @Ser0n-ath and @yashankxy who are also 4th year student at UofT CS. Can we take initiative to work on this issue and contribute to langchain?
Thanks 😊